Arch Linux

Please read this before reporting a bug:
https://wiki.archlinux.org/title/Bug_reporting_guidelines

Do NOT report bugs when a package is just outdated, or it is in the AUR. Use the 'flag out of date' link on the package page, or the Mailing List.

REPEAT: Do NOT report bugs for outdated packages!
Tasklist

FS#61458 - kernel 4.19 and later: scsi multi-queuing scheduler causes raid5 to stop

Attached to Project: Arch Linux
Opened by T. Baumann (dtbaumann) - Friday, 18 January 2019, 17:24 GMT
Last edited by Jan de Groot (JGC) - Friday, 31 May 2019, 06:50 GMT
Task Type Bug Report
Category Packages: Core
Status Closed
Assigned To No-one
Architecture x86_64
Severity Medium
Priority Normal
Reported Version
Due in Version Undecided
Due Date Undecided
Percent Complete 100%
Votes 0
Private No

Details

With all kernels after 4.19.1 more or less exactly 10 minutes after reboot md127_raid5 is blocked. Following that, the databases and all processes using the raid disks get stalled and the system get's unusable. A journal output after the update to 4.20.3 is attached with additional information. Following a suggestion (see https://bbs.archlinux.org/viewtopic.php?id=243520) I disabled the new scsi multi-queuing scheduler with

scsi_mod.use_blk_mq=0

and everything is back to normal. No problems at all with kernel 4.18.16.

Additional info:
* linux 4.20.3-1

Linux server 4.20.3-arch1-1-ARCH #1 SMP PREEMPT Wed Jan 16 22:38:58 UTC 2019 x86_64 GNU/Linux

* config and/or log files etc.
attached.

Steps to reproduce:

On my server:
+ Update to any kernel of the 4.19 or 4.20 series up to 4.20.3
+ Wait for 10 minutes
+ see error messages showing up in the logs and loose connection to mounted drives on the server

+ Downgrade to 4.18.6 or boot with scsi_mod.use_blk_mq=0
=> see no errors and enjoy a stable system
   arch-scsi-blk-mq-bug.txt (134 KiB)
This task depends upon

Closed by  Jan de Groot (JGC)
Friday, 31 May 2019, 06:50 GMT
Reason for closing:  Fixed
Comment by Jake Kreiger (Magali75) - Friday, 18 January 2019, 18:48 GMT
You should report it upstream https://bugzilla.kernel.org/ . Keep in mind that with Linux 5.0, scsi_mod.use_blk_mq=0 won't work anymore.
Comment by Brian Parsons (bparsons) - Saturday, 09 February 2019, 15:50 GMT
I have been having this problem as well.

A kernel bug was reported for this: https://bugzilla.kernel.org/show_bug.cgi?id=202353

A patch addressing is in 5.0-rc5. It can be applied to 4.19+: https://marc.info/?l=linux-kernel&m=154883900500866
Comment by loqs (loqs) - Tuesday, 19 February 2019, 03:51 GMT
queued for 4.20.11 https://git.kernel.org/pub/scm/linux/kernel/git/stable/stable-queue.git/commit/queue-4.20/blk-mq-fix-a-hung-issue-when-fsync.patch?id=a2649f44c7941e12b931d69b79ec73051ddde1a5
queued for 4.19.24 https://git.kernel.org/pub/scm/linux/kernel/git/stable/stable-queue.git/commit/queue-4.19/blk-mq-fix-a-hung-issue-when-fsync.patch?id=607bd65f4b2fe56644d0fdb6a2c2e0c586f2fb6b

Loading...